Differences

This shows you the differences between two versions of the page.

--- tips:model_selection_with_glmulti_and_mumin [2021/03/16 20:14] – Wolfgang Viechtbauer
+++ tips:model_selection_with_glmulti_and_mumin [2021/03/16 20:16] – Wolfgang Viechtbauer
@@ Line 205: / Line 205: @@
 </code>
-I rounded the results to 4 digits to make the results easier to interpret. Note that the table again includes the importance values. In addition, we get unconditional estimates of the model coefficients (first column). These are model-averaged parameter estimates, which are weighted averages of the model coefficients across the various models (with weights equal to the model probabilities). These values are called "unconditional" as they are not conditional on any one model (but they are still conditional on the 128 models that we have fitted to these data; but not as conditional as fitting a single model and then making all inferences conditional on that one single model). Moreover, we get estimates of the unconditional standard errors of these model-averaged values. These standard errors take two sources of uncertainty into account: (1) uncertainty within a given model (i.e., the standard error of a particular model coefficient shown in the output when fitting a model; as an example, see the output from the "best" model shown earlier) and (2) uncertainty with respect to which model is actually the best approximation to reality (so this source of variability examines how much the size of a model coefficient varies across the set of candidate models). The model-averaged parameter estimates and the unconditional standard errors can be used for multimodel inference, that is, we can compute z-values, p-values and confidence interval bounds for each coefficient in the usual manner.
+I rounded the results to 4 digits to make the results easier to interpret. Note that the table again includes the importance values. In addition, we get unconditional estimates of the model coefficients (first column). These are model-averaged parameter estimates, which are weighted averages of the model coefficients across the various models (with weights equal to the model probabilities). These values are called "unconditional" as they are not conditional on any one model (but they are still conditional on the 128 models that we have fitted to these data; but not as conditional as fitting a single model and then making all inferences conditional on that one single model). Moreover, we get estimates of the unconditional standard errors of these model-averaged values. These standard errors take two sources of uncertainty into account: (1) uncertainty within a given model (i.e., the standard error of a particular model coefficient shown in the output when fitting a model; as an example, see the output from the "best" model shown earlier) and (2) uncertainty with respect to which model is actually the best approximation to reality (so this source of variability examines how much the size of a model coefficient varies across the set of candidate models). The model-averaged parameter estimates and the unconditional standard errors can be used for multimodel inference, that is, we can compute z-values, p-values, and confidence interval bounds for each coefficient in the usual manner.
 ==== Multimodel Predictions ====